CodePrism
Builds a persistent code knowledge graph and exposes precise context to AI agents via MCP.
Tested on Linux
Use it
Linux x86_64
You need first
- Runtimes:
Python 3.12+
install codeprism-ai with uv tool
uv tool install codeprism-ai
Check that it works
codeprism --help 2>&1The output should contain “CodePrism”.
Add it to your agent
claude mcp add codeprism -- codeprism serve .JSON for Claude Desktop, Cursor and other clients
{
"mcpServers": {
"codeprism": {
"command": "codeprism",
"args": [
"serve",
"."
]
}
}
}Run codeprism serve . from your project folder, or add that command as an MCP server in your AI client.
What it does
- Build a lasting knowledge graph of your codeTested
How to use it
- Start the server in your project folder with
codeprism serve .. - From your MCP client, call the
index_projecttool with the path set to your project folder.
- Start the server in your project folder with
- Find callers, callees and impact of codeTested
How to use it
- Start the server in your project folder with
codeprism serve .. - Call the
get_calleestool with the file and the function name, for example <file> and <function>.
- Start the server in your project folder with
- Scan code changes for security issuesTested
How to use it
- Start the server in your project folder with
codeprism serve .. - Call the
scan_difftool with the original text, the proposed text and the file name.
- Start the server in your project folder with
- Give AI agents code context over MCPTested
How to use it
- Start the server in your project folder with
codeprism serve .. - Call the
get_contexttool with a file and a symbol name to get its context.
- Start the server in your project folder with
How far pickfit got
Linux x86_64
- Installs
- Connects
- 4 of 4 features work
- Works offline
Good for
- get structured context for a function with callers and callees
- scan proposed file changes for secrets before writing
- analyze impact and blast radius before refactoring
- track agent session reads and writes to avoid re-fetching
How pickfit tested it
Each run starts from a clean machine and follows the steps above. How pickfit verifies
Linux x86_64
An MCP client called one of its tools
- Replayed from scratch
- Works offline
- Checked on 30 Oct 2026
- Version 0.1.11
An MCP client handshake over stdio listed 21 tools, and a call to index_project returned a result.